Although stemmed femoral components often are used in revision total knee arthroplasty, no quantitative comparison of the relative stability of a femoral component with and without a stem has been performed previously. A radiostereophotogrammetric analysis was performed to determine the influence of stem extension on mechanical stability in a laboratory experiment. In addition, the contribution of impacted morselized bone graft used for reconstruction of bone defects to stability was determined. Ten fresh frozen distal femoral specimens, of which the bone mineral density was measured, were prepared to fit a cemented femoral component with an uncemented stem. A cyclic axial load of 750 N was applied to the medial part of the femoral component. The loading test was repeated after creation of a standard, unicondylar, uncontained medial bone defect, after reconstruction of the defect with impacted morselized bone graft, after disconnecting the stem from the component, and after removal of the morselized bone graft. A significant difference was found in rotation and translation of the femoral component with the stem and after its functional removal. The femoral component rotated significantly into varus and internal rotation, and the tip of the stem translated significantly laterally after disconnection of the stem. A linear correlation between bone mineral density in the femoral neck and the radiostereophotogrammetric motion data was absent, but a bone mineral density threshold level of 0.55 g/cm2 was present, below which stability decreased considerably. Morselized bone graft provided only a minor contribution to stability compared with a stem.